While there is a strong movement to develop new educational resources to bring students to the competencies represented by educational content standards, it is recognized that there are vast repositories of educational resources already developed that are suitable to address those competencies. However, these resources need to be indexed by national and state standards to make them accessible for teachers who are increasingly required to teach to certain educational standards (Diekema and Chen, 2005). In the early 1980s, a perceived a crisis in the American education system encouraged the creation of national standards by professional subject-area organizations such as the National Council for Teachers of Mathematics. These national standards aimed to clearly define what students in certain grade levels are expected to know in core subject areas (Ratvitch, 1995). Eventually all states followed suit and published their own educational standards, often using the national standards as a guideline.
Adding state standard information of every state to each resource is a large task, especially when done completely manually. Manual standard to standard alignment efforts (e.g. Align to Achieve) proved too difficult to maintain as the standards exist in all core subject areas; on national, state and local levels; and are revised regularly. Each set of standards utilizes discrete language, differing grade bands, distinct organizational structures and different levels of specificity in the coverage of a particular standard. To remedy these problems the Center for Natural Language Processing (CNLP) at Syracuse University has created a technology (Standards Alignment Tool – SAT) for automatically aligning state standards and national standards (Diekema et al., 2007).
This paper explores whether it is possible to exploit existing manual standards assignments by mining the groups of standards that have been assigned to a particular resource. In other words, rather than requiring explicit manual alignments between equivalent standards, this preliminary research is trying to use the assignment of standards as metadata to resources to determine which state standards might be equivalent. An increasing number of manual standards assignments is becoming available, possibly making this approach a viable and sustainable option. The ultimate goal of this research is to establish an automatic correlation between standards based on their shared occurrence. At the initial stage we are only considering groups of standards that were assigned to the same lesson plan. Eventually we’ll take into the co-occurrence statistics of standards across the entire corpus of lesson plans.